This change moves the vello encoding logic to a new crate under
crates/encoding. Combined with the `vello_shaders` crate, this enables
lightweight integration of the Vello pipelines into renderers that don't
depend on wgpu (or perhaps written in languages other than Rust).
The Scene/Fragment API currently remain the vello crate.
Naga traslates workgroup variable declarations to threadgroup
address-space entry-point parameters when generating MSL. Metal API
validation requires that the memory sizes for these parameters be set
explicitly by calling setThreadgroupMemoryLength:index on the
MTLComputeCommandEncoder.
The crate now calculates the required memory size for global workgroup
variables that are accessed by the entry point and provides them
alongside the binding list. This is abstracted separately from the
binding list.
While the current usage that we're aware of is limited to Metal, this
information is being provided as part of the generic ComputeShader type
instead of a MSL-specific type, as the information itself is computed
from the parsed WGSL IR and not specific to Metal.
Using vector graphics for the image doesn't make a huge amount of sense. The flower photo is by Raph and happily licensed to anyone who wants to use it.
This allows the graph to display at a reasonable scale in the face of
fluctuations and a max recorded sample that is much larger than the
current average.
The Auto* modes should have wider compatibility as they implement
fallback behavior based on what the platform supports. This also means
that on web "vsync off" will likely be incorrect.
Added a module for frame time statistics and UI layer that displays
the average, minimum, and maximum frame time alongside FPS. The UI
can be toggled by pressing the `S` key.
* Add comment about naughty ptr2int cast
* Change Option to anyhow::Result in test scene ImageCache
* Replace magic constant with pixel stride from selected format
* reorganize GlyphCache::get_or_insert() to use the HashMap entry API
* change DrawGlyphs::hint() parameter name from `yes` to `hint` for clarity
* change Layout::path_data() to return bytes instead of [f32; 2]. Path segments could be encoded as i16, making this incorrect
